This revised competing continuation application (MH061566) describes our past progress and continuing studies to test a sleep neuroscience model of depression. In the prior funding period, in relation to healthy subjects, our sleep neuroimaging studies in depression have identified a ventral limbic and paralimbic neural system, including the amygdala and subgenual anterior cingulate cortex (ACC), that activates greater during REM sleep compared with waking, and a dorsal executive neural system, including the dorsolateral prefrontal cortex and dorsal ACC, that shows less deactivation during NREM sleep in relation to waking. These two neural systems have independently been reported to play significant roles in emotional behavior, in general, and in the abnormal emotional behavior of depression. The overall goal of the current proposal is to link the sleep neuroimaging findings in depression with this developing cognitive and affective neuroscience literature in depression. We propose two separate mediational models to explain the relationships between sleep and emotional behavior in depressed patients. First, we hypothesize that the effects of depression on increasing ventral neural system function in REM sleep is partially mediated by an increase in emotional responsivity as measured by ventral neural system (amygdala/subgenual ACC) fMRI responses to an emotional stimuli. Second, give a substantive literature supporting a role for NREM sleep in restoring prefrontal cortex function, we hypothesize that the effects of depression on decreased prefrontal cortex and dorsal ACC function, as measured by fMRI cognitive probes, is in part mediated by the lack of restoration of the prefrontal cortex and dorsal ACC during NREM sleep in depressed patients. To test these mediational models, we will assess cerebral metabolic responses to waking, REM and NREM sleep probes using [18F]-FDG PET and cerebral blood flow responses to affective and cognitive challenges using fMRI in 35 depressed and 35 age- and gender-matched and IQ equivalent healthy subjects. Repeated measures MANCOVAs will test group x condition interactions in the sleep, affective and cognitive neuroimaging activation paradigms. Multiple linear regressions will be used to test the mediational models above. Findings from this study are expected to significantly advance our understanding of the neuroscience of depression.